enricobuehler
98e68a49a7
Security review 2026-08-25, 58 confirmed findings across host, console, clients and supply chain. Nearly every serious one is a documented boundary whose code stopped enforcing what its comment promised — so where the two disagreed, the comment won and the code was made to match, and where it could not be, the comment was corrected instead. Critical — a console session cookie alone reached code execution: every pairing route rode the generic catch-all with the operator's admin bearer attached. Arming, approving and PIN submission now sit behind the console password like the other trust-root routes, and the armed PIN is returned once in that gated response instead of riding a 1 s status poll. High — the plugin lane no longer reads the unredacted log ring (which carried the webhook credentials the /hooks carve-out exists to withhold); hook lines log an origin and a short id, never a URL or a command line; a plugin-reported pid is held to procscan's start-time floor before the SYSTEM host will signal it; ClipOffer is gated on the live grant mask, so a revoked guest loses the host clipboard in both directions; ENet refuses connects with no live launch instead of letting LAN peers squat all four slots; Windows secrets are born with their DACL applied rather than world-readable; the sysext feed binds FEED and a monotonic SERIAL inside the signed bytes; privileged_field allowlists the host-resolved launch kinds so a new kind is privileged by default; five parser panics reachable from one malformed NALU are range-checked; release-signing jobs pin bun, sccache and actions by checksum/SHA; h2 -> 0.4.19 (RUSTSEC-2026-0258). Deep links only auto-dial by stable record id now — a display name or an address gets a confirmation on every client. The Apple identity key moves to ThisDeviceOnly so it stops riding encrypted backups. pf-vdisplay stops routing session identity through the process environment: the injector backend threads through a typed slot, so per-batch getenv no longer races a per-session setenv. The four remaining writes have no in-repo readers and are documented as such; the SAFETY proof that claimed ENV_LOCK made them sound is gone. Verified: cargo clippy --workspace --all-targets --locked -D warnings and cargo fmt --all --check clean in the CI image; web builds, tsc --noEmit clean, 22/22 server tests; Swift debug+release + 26/26, Kotlin :kit 7/7. Not fixed, deliberately: the plugin token can still mint command execution (the plugin launch kind exists so a plugin names a command the host runs — per-plugin identity does not change that, and the runner is one process hosting fibers, so there is nowhere to hang a credential); the shared plugin-UI origin; the rollback Authenticode publisher pin (Azure mints a fresh leaf per request, and the signer subject is not in the tree). Each is now described accurately where it lives instead of being claimed closed.
12 KiB
title, description
| title | description |
|---|---|
| Updating the Host | How to see when a newer Punktfunk host is available — the web console's update card — and the update command for every install method. |
The web console tells you when a newer host is out. The Host page has an Updates card
showing the version you run, the channel you follow (stable or canary), how this host was
installed, and — once a newer release exists — the exact command that updates it. The
"update available" state also fires an update.available event on the host
event stream, and a successful update fires update.applied (with from
and to) once the host is back up, so hooks and scripts can react to both.
Your channel comes from the repository this host installs from — see Release Channels for what each track means and how to move a host between them. The Updates card never switches channels for you.
The check is a small signed manifest the host fetches from the Punktfunk release feed and verifies against keys built into the host itself — a tampered or replayed feed is rejected, and the console says when a check failed rather than silently showing stale facts.
Updating, per install method
The console shows the right one of these automatically; for reference:
| How you installed | How to update |
|---|---|
| Windows installer | download the newer punktfunk-host-setup-<version>.exe from the releases page and run it — it upgrades in place, keeping your settings, console password and paired devices |
| Windows via winget | winget upgrade unom.PunktfunkHost — register the Punktfunk source once first, in an elevated terminal: winget source add -n punktfunk https://winget.punktfunk.unom.io -t Microsoft.Rest |
| Ubuntu (apt) | sudo apt update && sudo apt install --only-upgrade punktfunk-host |
| Fedora (dnf) | sudo dnf upgrade punktfunk |
| Bazzite sysext (recommended) | sudo punktfunk-sysext update |
| Bazzite / Fedora Atomic rpm-ostree layer | see below — rpm-ostree upgrade alone is not enough (staged — reboot to finish) |
| Arch / CachyOS (pacman) | sudo pacman -Syu (a normal full system upgrade) |
| Steam Deck (on-device build) | bash ~/punktfunk/scripts/steamdeck/update.sh --pull |
| NixOS (flake) | nix flake update punktfunk in your flake directory, then rebuild your system |
winget carries stable releases only. A Windows host on the canary channel updates by running the
canary installer again — …/generic/punktfunk-host-windows/canary/punktfunk-host-setup.exe, see
Release Channels.
rpm-ostree layer: rpm-ostree upgrade is not enough
rpm-ostree upgrade upgrades the base image and only re-resolves layered packages when the
base actually changes — so on a base that sits still (a pinned tag, a paused rebase) it keeps
reporting "No updates available" while newer Punktfunk RPMs sit in the repo. Force it to
re-resolve just the Punktfunk layer, removing and re-adding the same names in one transaction:
sudo rpm-ostree refresh-md --force
sudo rpm-ostree update \
--uninstall punktfunk --uninstall punktfunk-web \
--install punktfunk --install punktfunk-web
systemctl reboot
Name only the packages you actually layered — rpm-ostree status lists them. The new version is
staged; it activates on the next boot.
The re-resolve picks the highest version across every enabled /etc/yum.repos.d/punktfunk*.repo,
so if the canary repo is enabled alongside the stable one, canary wins and the box quietly tracks
canary — enable exactly the channel you want (see Release Channels). And if this
box runs the Bazzite sysext, the sysext shadows any layered copy: update with
sudo punktfunk-sysext update instead.
Bazzite sysext: channels, rollback and rebases
sudo punktfunk-sysext update fetches and merges the newest build of whatever channel the last
install wrote to /etc/punktfunk-sysext.conf (status shows channel, installed and latest
version). To switch channel, re-run the install: sudo punktfunk-sysext install --channel canary
(or --channel stable) — update takes no channel flag. To be able to go back to a build that
worked, keep a copy of the image before updating, and re-install that file afterwards:
sudo cp /var/lib/extensions/punktfunk.raw ~/punktfunk-known-good.raw # before updating
sudo punktfunk-sysext install --from-file ~/punktfunk-known-good.raw # to go back to it
- After a Bazzite major rebase (Fedora 43 → 44) the old image refuses to load rather than run
against mismatched system libraries — run
sudo punktfunk-sysext updateonce and it fetches the image built for the new base. - If it refuses the feed.
refusing to install from an unsigned feedmeans that Fedora major's feed predates signing; it gets sealed on the next publish.signed but UNBOUNDis the same kind of thing one step on — a feed sealed before manifests carried their# FEED/# SERIALheader — and it too is fixed by the next publish. To install from either anyway, accepting unauthenticated images,sudo env PUNKTFUNK_SYSEXT_ALLOW_UNSIGNED=1 bash punktfunk-sysext.sh install. The other three messages are not the same thing — don't install; re-download the script and try again:the feed's SHA256SUMS is NOT signed by packages@unom.io,this manifest was signed for the feed '…'(another channel's or another Fedora major's feed served as yours), andmanifest serial … is older than the last accepted …(a real but superseded manifest replayed). - If it refuses to downgrade.
the feed's newest image (…) is OLDER than the installed …means the feed lost a build —updatewon't walk you back onto a superseded release without being asked. For a deliberate rollback,sudo PUNKTFUNK_SYSEXT_ALLOW_DOWNGRADE=1 punktfunk-sysext update, or re-install the image file you kept above, which never consults the feed. - The feed's checksum manifest is OpenPGP-signed by packages@unom.io (key
AF245C506F4E4763, the same one that signs the RPMs) andpunktfunk-sysextverifies it against a key baked into the script, so it needsgpgon the box. The manifest also names the feed it was signed for and carries a publish serial, both inside the signed bytes, so a signed manifest from another channel — or an older one put back — can't be replayed at yours.
Restart after a Linux package update
Restart the host to pick up the new binary:
systemctl --user restart punktfunk-host
If the update also brought a new punktfunk-web (the console — a separate package and a separate
user service), restart that too, and first; the page blinks and reconnects:
systemctl --user restart punktfunk-web
systemctl --user restart punktfunk-host
A pacman -Syu upgrades every installed Punktfunk package, so it always needs both. The apt and
dnf commands above name one package, so they usually don't — to move the console and the host
together, name both: sudo apt install --only-upgrade punktfunk-host punktfunk-web (apt) or
sudo dnf upgrade punktfunk punktfunk-web (dnf). If you enabled the plugin/script runner, it is
a third unit: systemctl --user restart punktfunk-scripting.
(The Windows installer restarts the service itself; punktfunk-sysext update prints the same
restart hint when it's needed.)
One-click updating (Windows)
On a Windows host the card shows an Update now button instead of a command. It asks for the console password again (a saved login alone can't restart your host), then the host downloads the installer, verifies it against the signed release manifest and its code signature, and runs it silently — the service restarts at the end and the page reconnects by itself. If a stream is live you're warned first: updating drops it.
Every attempt leaves a result in the card (and an installer log under
C:\ProgramData\punktfunk\logs\update-<version>.log — readable only by Administrators and
SYSTEM, so open it from an elevated PowerShell) — including across the restart, so a
failed update is never silent.
If the newly installed host crash-loops, the service puts the previous installer back on its own
(the last two are kept) and says so in the card — you end up on the version you started from, not
on a dead host. That rollback writes
C:\ProgramData\punktfunk\logs\update-rollback-from-<version>.log.
One-click updating (Linux — opt-in)
The apt, dnf, Bazzite-sysext, and rpm-ostree installs can one-click update too, via a small root
helper the packages ship (pf-update + a punktfunk-update.service oneshot). It's off until you
opt in — a web button that ends in root deserves an explicit decision:
sudo usermod -aG punktfunk-update $USER # takes effect within a minute — no re-login needed
The console re-checks group membership every minute, so the Update now button replaces the opt-in hint on its own.
That group membership is the entire grant — a polkit rule lets its members start exactly that one service, whose only job is "run this system's normal package update for the Punktfunk packages, then prove the new binary runs". The button never chooses versions or URLs; your package manager's own signed repositories stay the source of truth. The manual command always keeps working.
Per method: on rpm-ostree the update is staged and the card says so — reboot to finish (the
console never reboots your machine). On Arch/pacman the button additionally requires
PACMAN_FULL_SYSUPGRADE=1 in /etc/punktfunk/update.conf, because the only safe pacman update is
a full pacman -Syu — we won't run a partial upgrade. After a successful update the host restarts
itself and the page reconnects.
The Steam Deck on-device build gets the button too, with no opt-in (it's your own user's
install, no root involved): it runs the same update.sh rebuild the docs describe, which compiles
on the Deck — expect it to take a while; the card shows progress and the log lands in
~/.config/punktfunk/logs/update-steamos.log.
Updating a client
This page is about the host. Clients update on their own tracks — flatpak update on Linux, your
normal apt upgrade / dnf upgrade for the packaged Linux client, a newer .dmg dragged over the
old app on macOS, TestFlight on iPhone/iPad/Apple TV, Google Play on Android, and the Decky panel's
own Update button on a Steam Deck. The per-platform table is in
Install a Client → Keeping a client up to date.
A host and a client don't have to be on the same version, but keeping them close is least
surprising.
Turning the check off
The check contacts git.unom.io (the Punktfunk forge) and nothing else, and sends nothing but a
normal download request. To stop the host checking, add this line to its host.env:
PUNKTFUNK_UPDATE_CHECK=0
host.env lives at ~/.config/punktfunk/host.env on Linux and
%ProgramData%\punktfunk\host.env on Windows — see Configuration. Then
restart the host: systemctl --user restart punktfunk-host on Linux, or restart the Punktfunk
host service on Windows. The card then shows checks as disabled; everything else keeps working.
PUNKTFUNK_UPDATE_APPLY=0 in the same file removes the Update now button on any host,
Windows or Linux; the card shows the manual command instead.
If the card says the feed is stale
"Feed hasn't changed in over 45 days" means checks succeed but nothing new arrives. Usually no
release happened for a while; if the releases page
shows something newer than the card does, something between this host and the feed is pinning old
data — look at proxies or DNS on the way to git.unom.io. That comparison only works on a
stable host: the releases page is stable-only, so a canary host being "behind" it means
nothing.
Next
- Something went wrong during or after an update? Troubleshooting.
- Want to remove Punktfunk instead, or find out what an uninstall leaves behind? Uninstalling.